Institutional review board approval was secured in June 1996. The study was performed from June through August 1996. Eligible patients were aged 18 years or older and were born in a country outside the United States and its territories. Patients born in this country and in US territories (Puerto Rico and Guam) were excluded because of their low prevalence of torture.
A nonconsecutive convenience sample of patients was drawn from patients attending the internal medicine ambulatory care clinic of a large, urban, municipal medical center in New York City. This clinic registered 9,547 visits among 2,700 unique patients during the study period. Two trained research assistants enrolled participants at 75% of clinic sessions during the study period. First, they examined the clinic registration and identified all eligible patients. About 50% of the registered patients were eligible. Second, they approached those patients who were in the clinic waiting room. The large number of patients in this clinic at this medical center made it logistically impossible to reach most persons in a timely manner. About 15% of patients were approached (selected only on the basis of availability). Third, the research assistant asked the patient to participate in a study about “the health care needs of foreign-born patients.” No reimbursement or incentives were offered for participation in the study. One of the research assistants was bilingual in Spanish. Hospital translators were used for participants who spoke French, Bengali, Russian, Polish, or Chinese (Mandarin and Cantonese dialects). To encourage participation and disclosure and to ensure safety, all participants were interviewed in a private office without friends or family.
The research assistants verbally administered the interview using the Detection of Torture Survivors Survey. The survey instrument has been validated against a standardized instrument and blinded, clinical interview for screening culturally heterogeneous clinical populations for exposure to torture.9
The primary outcome was a self-reported history of torture, as determined by the survey. The Detection of Torture Survivors Survey uses the definition of torture of the World Medical Association's Declaration of Tokyo: “the deliberate, systematic or wanton infliction of physical or mental suffering by one or more persons acting alone or on the orders of any authority, to force another person to yield information, to make a confession or for any other reason.”10
All survivors of torture were asked whether the physicians who were treating them knew about their reported history of torture. Also, the investigators reviewed participants' medical records and interviewed the patients' primary care physicians to determine whether the physicians either had made notes on or knew about their patients' reported history of torture. The participants' presenting complaints and their treatments were not systematically recorded.
Participants whose histories of torture fit the criteria in the Detection of Torture Survivors Survey were interviewed by clinicians who were expert in the evaluation of torture survivors. The expert interviewers came from the medical and psychology staff of the Bellevue—New York University Program for Survivors of Torture. Clinical assessment was chosen because research indicates that clinical interviews have good diagnostic accuracy for defining exposure to torture.11
The clinical assessment included taking a detailed history of medical and psychiatric problems and of trauma. The history of trauma was correlated with what is known about the conditions in the country where the trauma was said to have taken place. To further verify each participant's reported exposure to torture would require a full physical examination and possibly ancillary testing.8
Such an examination can be both time-consuming and retraumatizing for a survivor.12
We did not try to obtain complete forensic verification of the reported torture because we decided it was unethical.
We carried out a univariate analysis that included means, standard deviations (SDs), and 95% confidence intervals (CIs). Bivariate analyses for differences in means were computed by the Student t test. Differences in proportions were computed using a χ2 test.